Free access
Features:
- Functional descriptive analysis
- Functional principal component analysis
- Functional principal component linear regression
- Functional principal component logit regression
- Functional descriptive analysis
- Functional principal component analysis
- Functional principal component linear regression
- Functional principal component logit regression
You only need to have as much two text files in csv (comma separated value) format to start to use the application: the file with the domain points where curves have been observed, and the file with the discrete observations of the curves in the previous sampling points. Upload your files and be ready to enjoy a different way to analize your data
Select the functional basis representation that best fits to your discrete data. You only have to select a type of basis and the dimension to use. You don't need to know anything about the mathematical details, the application offers you different possibilities and you can visually check the result before deciding. You can change the different possibilities as many times as necessary, and if the final result is not satisfactory, you will be able to return and try different options.
Choose the functional data analysis methodology that best suits your objective among the ones proposed (exploratory analysis, dimension reduction, linear fit and prediction or binary fit and prediction). Get your results and draw conclusions (see details of each method in the documentation of the application).
Functional data analysis is a statistical methodology in which the input information used to obtain the output results, is a set of curves (mathematical functions). Imaging, for example, a variable whose values depend continously on time, it would be better to use the information of such a variable at any time point than using it only at a finite set of fixed time points. All FDA methods take advantadge of using the complete curve instead of a reduced set of discrete observations that could lead to loose relevant information. This way FDA takes into account relevant features of curves (continuity, derivability, smoothness, maxima and minima, etc) to provide more accurate knowledge for making optimal decisions based on the results of statistical analysis
In spite of FDA methodologies use the complete curve to get their aims, it is technically impossible to record the mathematical function itself in the observation process of any functional data set. At most we could register dicretized observations of such curves at a dense set of points of their domain (in many cases we only have sparse observations at diffeten sampling points that could even be different for the sample units). So the process in all FDA methods and also in this application begins by uploading the discrete observations of curves. Different situations are taken into account: all curves observed at the same domain points, different curves observed at different domain points... After uploading the data, the application guides you to get the mathematical form of your curves. Finally, you will be able to select the statistical method that best suits your aim and get the results of the selected analysis